PDF-creating software that can "re-use" image data...? (PDF "Image XObjects")

Question

I need to create a PDF file automatically (preferable via a batch process / command-line, the optimum would be something XSLT/SAXON-based).

The PDF will have many pages (ca. 500), and each page contains nothing but a set of (partially overlapping) images (in scalable format, e.g. SVG or WMF).

But there is only a basic set of ~20 images used for this, and the various pages differ one from another only in the particular choice out of these 20 images.

The problem: If I use XSL-FO and feed it into the Apache FOP then the result is a PDF of approx. 300 MB file-size, because the image data is fed multiple times into the PDF.

Now I am looking for a possibility to create a PDF in a way such that the image data of these 20 images are present only once in the PDF, and each PDF-page uses only "links"/references to this set of image data (so that the resulting PDF-file has only ~2MB file-size).

I know that the PDF-language as such features such references, but of course, I am not interested to build a PDF-file directly, "low Level" ;-)

Any help/ideas would be highly appreciated! Thanks a lot!

Answer 1

Report Lab has both open source software to do this sort of thing, (from python), and paid for Report Lab Plus tool that you can use via their Report Markup Language (RML).

Answer 2

You can easily accomplish this with PDFTron PDFNet.

Image shared_img_object = Image.Create(doc, input_path + "peppers.jpg");
// then per page add image this way

Element element = builder.CreateImage(shared_img_object, new Matrix2D(sx, 0, 0, sy, dx, dy));
writer.WritePlacedElement(element);

See the AddImage or ElementBuilder samples for more details.

Answer 3

Well, after all, I decided in favor of a well-known well-established tool, "(plain) pdfTeX". Its \pdfximage and \pdfrefximage macros offer exactly what I needed. (Of course, pdfTeX cannot handle SVG, but converting my SVG images to a set of PDF files was not a problem, especially since this needed to be done only once.) So, as a result, I have an XSLT now that converts my input XML to (plain) pdfTeX, and the whole conversion works amazingly fast on modern PCs.